An Image Encryption Algorithm Based on HNN with Memristor

This paper proposes an innovative approach to designing image encryption hardware by leveraging the Negative-resistance Memristor-based Hopfield Neural Network (NMHNN) model. In this method, the conventional Hopfield Neural Network (HNN) is modified by substituting one of its synaptic weights with a negative-resistance memristor model. This modification demonstrates enhanced complex dynamics while maintaining a simplified structure. The NMHNN generates a chaotic sequence, which serves as a key for image encryption and decryption through a confusion–diffusion process. As a result, this approach significantly improves the efficiency of both image encryption and decryption processes. To validate the practical feasibility of this memristor-based neural network for encryption, a 3-neuron HNN circuit with one [Formula: see text] memristor is constructed and tested. The hardware experiment demonstrates strong resilience against statistical analysis and entropy attacks. Notably, the operational efficiency of the proposed method is demonstrated to be 23 times greater than that of the Advanced Encryption Standard (AES), highlighting its substantial effectiveness and potential for hardware implementation in the field of image encryption.